Small Single-Stranded DNA Structure Prediction.

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Small single-stranded DNAs (ssDNAs) are crucial therapeutic agents, but their 3D structures are vital for development.
• Experimental structure determination is costly and time-consuming, limiting available structural data.
• Accurate computational methods are needed to predict ssDNA structures efficiently.

Purpose of the Study:

• To evaluate the performance of AlphaFold 3 (AF3) against traditional methods for ssDNA structure prediction.
• To assess the impact of molecular dynamics (MD) simulations on AF3-generated ssDNA structures.
• To provide guidance for using AF3 and MD for ssDNA structure sampling.

Main Methods:

• Comparison of a traditional multistep ssDNA structure prediction approach with AlphaFold 3 (AF3).
• Evaluation using a dataset of 149 ssDNAs across six structural motifs.
• Assessment of AF3 performance with and without subsequent molecular dynamics (MD) simulations.

Main Results:

• AF3 demonstrated higher efficiency and accuracy than the traditional method, achieving a 38% success rate (GDT=1).
• AF3 faced challenges with longer sequences and complex motifs like G-quadruplexes and junctions.
• MD simulations significantly improved the quality of AF3-predicted ssDNA structures.

Conclusions:

• AF3 is a valuable tool for ssDNA structure prediction, though limitations exist for complex cases.
• Combining AF3 with MD simulations offers a robust strategy for accurate ssDNA structure determination.
• These findings facilitate the use of computational methods in advancing ssDNA-based therapeutics.